Make InferCtxt::can_eq and InferCtxt::can_sub return booleans

[compiler-errors]

Nobody matches on the result, nor does the result return anything useful...

[rustbot]

r? @lcnr

(rustbot has picked a reviewer for you, use r? to override)

[compiler-errors]

After #107940 lands, this these functions should additionally use TyCtxt::check_tys_might_be_eq when the new solver is enabled, to check for deep type equality with lazy norm... but I'll leave that up to another PR.

[BoxyUwU]

I feel like if !self.infcx.can_eq(a, b) { is less clear than if self.infcx.can_eq(a, b).is_err(). Could we just make the return type Result<(), ()>. if we start handling nested obligations we wont be able to always return a TypeError anyway so getting rid of it would have to happen

[compiler-errors]

The answer to a "can" question being a boolean makes pretty good sense to me, and !self.can_do_thing() is pretty clearly "can't do thing".

We also have FnCtxt::can_coerce which has a handful of !self.can_coerce calls around the codebase already.

[lcnr]

Ignoring for a moment that can_eq is kinda bad as a whole because it ignores nested obligations causing stuff like lazy norm to end up changing behavior I am in favor of this change.

Definitely in favor of yeeting TypeError from can_eq because I can't think of a case where you should use that TypeError to actually do anything 😅

wrt TyCtxt::check_tys_might_be_eq and lazy norm I don't yet have an opinion. I slightly prefer returning bool over result here.

[lcnr]

@bors r+

[bors]

📌 Commit eb286dd has been approved by lcnr

It is now in the queue for this repository.